Ryan M. Robinson

High Specific Power Actuators for Robotic Manipulators

Recent advancements in actuator technology suggest that the implementation of reliable, high power-to-weight ratio pneumatic actuation systems is now possible for robotic platforms. Existing robotic manipulator arms for casualty extraction and patient placement use hydraulic actuation, whereas related robotic prosthetic devices typically use heavy actuator motors. We have developed an alternative solution that employs pneumatic artificial muscles (PAMs). The goal of this study is to identify requirements for a lightweight, high-force robotic manipulator, design the system for heavy lifting capability, and assemble a prototype arm. Following characterization and comparison of different-sized PAM actuators, a proof-of-concept manipulator was constructed. A quasi-static model for the PAM actuators was applied to the system, which includes the Gaylord force, as well as non-linear elastic energy storage. Experimental testing was performed to measure the joint torque and dynamic response of the manipulator, and to validate the model.

Variable Recruitment Testing of Pneumatic Artificial Muscles for Robotic Manipulators

This paper investigates the orderly recruitment of pneumatic artificial muscles for efficient torque production in a robotic manipulator. Pneumatic artificial muscles (PAMs) are arranged in a parallel bundle, and independently-controlled “motor units” are employed to imitate the structure and function of human skeletal muscle. Simulated cycling tests are conducted on a model of the robotic manipulator to quantify the benefits of variable recruitment, and experimental testing is performed to validate the simulated predictions. Results reveal a distinct relationship between recruitment and system efficiency. Key factors influencing the value of a variable recruitment strategy include nonlinear PAM bladder elasticity, pneumatic losses, and dissipative forces in the robotic joint. Recruitment guidelines are proposed to maximize efficiency over a range of payload masses. The potential challenges associated with maintaining smooth motion control during discrete transitions in recruitment are also identified and discussed.

Nonlinear Control of Robotic Manipulators Driven by Pneumatic Artificial Muscles

Lightweight, compliant actuators are particularly desirable in safety-conscious robotic systems intended for interaction with humans. Pneumatic artificial muscles (PAMs) exhibit these characteristics and are capable of higher specific work than comparably sized hydraulic actuators and electric motors. However, control of PAM-actuated systems has proven difficult due to the highly nonlinear nature of the actuators and the pneumatic systems driving their actuation. This study develops and investigates the performance of three advanced control strategies-sliding mode control, adaptive sliding mode control, and adaptive neural network (ANN) control-each containing a distinct level of a priori model knowledge, to enable smooth and accurate motion tracking of a single degree-of-freedom PAM-actuated manipulator. Originally developed by J.-J. Slotine and R.M. Sanner, the specific controllers employed in this study are significantly modified for application to pneumatically actuated open-chain manipulators with complex nonlinear dynamics. The two adaptive controllers are updated online and require no pretraining step. Several experiments are performed with each controller to evaluate and compare closed-loop tracking performance. Results highlight the dependence of a preferred control strategy on the level of model completeness and quality, and suggest that in most PAM-actuated manipulator scenarios, the ANN controller is preferable because it does not require a model of the pneumatic system or joint mechanism design, which can be difficult and time consuming to characterize, and is robust to changes in PAM actuator characteristics (due to fatigue or replacement).

Dynamic belief fusion for object detection

A novel approach for the fusion of heterogeneous object detection methods is proposed. In order to effectively integrate the outputs of multiple detectors, the level of ambiguity in each individual detection score is estimated using the precision/recall relationship of the corresponding detector. The main contribution of the proposed work is a novel fusion method, called Dynamic Belief Fusion (DBF), which dynamically assigns probabilities to hypotheses (target, non-target, intermediate state (target or non-target)) based on confidence levels in the detection results conditioned on the prior performance of individual detectors. In DBF, a joint basic probability assignment, optimally fusing information from all detectors, is determined by the Dempsters combination rule, and is easily reduced to a single fused detection score. Experiments on ARL and PASCAL VOC 07 datasets demonstrate that the detection accuracy of DBF is considerably greater than conventional fusion approaches as well as individual detectors used for the fusion.

Linking Porosity and Tortuosity to the Performance of a Magneto-Rheological Damper Employing a Valve Filled With Porous Media

The effects of porous media selection on the performance of a porous-valve-based magnetorheological (MR) bypass damper are evaluated. Important media parameters affecting the damper performance (i.e., porosity and tortuosity) are identified using flow analysis. The relationship between the controllable force of the damper and porous valve characteristics is studied for packed beds of cylindrical rods and spheres. Damper performance is compared between the different porous media configurations, with primary attention given to the maximum controllable force and damping coefficient.

A design strategy for magnetorheological dampers using porous valves

To design a porous-valve-based magnetorheological (MR) damper, essential design parameters are presented. The key elements affecting the damper performance are identified using flow analysis in porous media and an empirical magnetic field distribution in the porous valve. Based on a known MR fluid, the relationship between the controllable force of the damper and the porous valve characteristics, i.e. porosity and tortuosity, is developed. The effect of the porosity and tortuosity on the field-off damping force is exploited by using semi-empirical flow analysis. The critical flow rate for the onset of nonlinear viscous damping force is determined. Using the above design elements, an MR damper using by-pass porous valve is designed and tested. The experimental damper force and equivalent damping are compared with the predicted results to validate this design strategy.

Human-autonomy sensor fusion for rapid object detection

Human-autonomy sensor fusion is an emerging technology with a wide range of applications, including object detection/recognition, surveillance, collaborative control, and prosthetics. For object detection, humans and computer-vision-based systems employ different strategies to locate targets, likely providing complementary information. However, little effort has been made in combining the outputs of multiple autonomous detectors and multiple human-generated responses. This paper presents a method for integrating several sources of human- and autonomy-generated information for rapid object detection tasks. Human electroencephalography (EEG) and button-press responses from rapid serial visual presentation (RSVP) experiments are fused with outputs from trained object detection algorithms. Three fusion methods-Bayesian, Dempster-Shafer, and Dynamic Dempster-Shafer-are implemented for comparison. Results demonstrate that fusion of these human classifiers with computer-vision-based detectors improves object detection accuracy over purely computer-vision-based detection (5% relative increase in mean average precision) and the best individual computer vision algorithm (28% relative increase in mean average precision). Computer vision fused with button press response and/or the XDAWN + Bayesian Linear Discriminant Analysis neural classifier provides considerable improvement, while computer vision fused with other neural classifiers provides little or no improvement. Of the three fusion methods, Dynamic Dempster-Shafer Theory (DDST) Fusion exhibits the greatest performance in this application.

Quasi-static nonlinear response of pneumatic artificial muscles for both agonistic and antagonistic actuation modes:

Pneumatic artificial muscles are actuators known for their low weight, high specific force, and natural compliance. Employed in antagonistic schemes, these actuators closely mimic biological muscle pairs, resulting in applications for humanoid and other bio-inspired robotic systems. Such systems require precise actuator modeling and control in order to achieve high performance. In the present study, refinements are introduced to an existing model of pneumatic artificial muscle force-contraction behavior. The force-balance modeling approach is modified to include the effects of non-constant bladder thickness and up to a fourth-order polynomial stress–strain relationship is adopted in order to accurately capture nonlinear pneumatic artificial muscle force behavior in contraction and extension. Moreover, the polynomial coefficients of the stress–strain relationship are constrained to vary linearly with pressure, improving the ability to predict behavior at untested pressure levels while preserving model accuracy at tested pressure levels. Lastly, a detailed geometric model is applied to improve force predictions, particularly during pneumatic artificial muscle extension. By modeling the deformation shape of the actuator ends as sections of an elliptic toroid, pneumatic artificial muscle force predictions as a function of strain are improved. These modeling improvements combine to enable enhanced model-based control in pneumatic artificial muscle actuator applications.

Dynamic inverse models in human-cyber-physical systems

Human interaction with the physical world is increasingly mediated by automation. This interaction is characterized by dynamic coupling between robotic (i.e. cyber) and neuromechanical (i.e. human) decision-making agents. Guaranteeing performance of such human-cyber-physical systems will require predictive mathematical models of this dynamic coupling. Toward this end, we propose a rapprochement between robotics and neuromechanics premised on the existence of internal forward and inverse models in the human agent. We hypothesize that, in tele-robotic applications of interest, a human operator learns to invert automation dynamics, directly translating from desired task to required control input. By formulating the model inversion problem in the context of a tracking task for a nonlinear control system in control-a_ne form, we derive criteria for exponential tracking and show that the resulting dynamic inverse model generally renders a portion of the physical system state (i.e., the internal dynamics) unobservable from the human operators perspective. Under stability conditions, we show that the human can achieve exponential tracking without formulating an estimate of the systems state so long as they possess an accurate model of the systems dynamics. These theoretical results are illustrated using a planar quadrotor example. We then demonstrate that the automation can intervene to improve performance of the tracking task by solving an optimal control problem. Performance is guaranteed to improve under the assumption that the human learns and inverts the dynamic model of the altered system. We conclude with a discussion of practical limitations that may hinder exact dynamic model inversion.

Passive switched system analysis of semi-autonomous systems

While autonomous capabilities have proliferated across a wide range of commercial and domestic applications, some tasks require intermittent aid from a human operator. Guaranteeing the safety of these intermittently-teleoperated systems requires stability guarantees that hold in the presence of switching. In this paper, we consider the problem of controlling a robotic vehicle using both a human controller and an autonomous controller. The strategy is to allow the human operator to switch between manual control and autonomous control as needed. The feedback loop is analyzed and shown to be stable using a notion of passivity from nonlinear system analysis. Finally, an example is provided to demonstrate the approach.